| Filename | Latest commit message | Latest commit date |
|---|---|---|
e6400ca053
Based on evaluation of both prototype implementations (v2/prototype/shadcn and v2/prototype/mantine), recommend Shadcn for the following reasons: - Component ownership: Components copied to codebase, not npm dependency - Tailwind CSS integration: Utility-first approach with purged bundles - Radix UI foundation: WCAG 2.1 compliant accessibility primitives - Simpler theming: CSS variables vs React context re-renders - Better DX: Source code visible and editable in project Includes comparison table and tradeoffs acknowledged. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> |
||
| .. | ||
| README.md | ||
| v1_problems.md | ||
| v2_scope.md | ||
| v2_tech_stack.md | ||
| v2_ux.md | ||
Inspector V2 Brief
Brief | V1 Problems | V2 Scope | V2 Tech Stack | V2 UX
Table of Contents
- Motivation and Context
- Track Protocol Conformance
- Improve Maintainability
- Provide Extensibility
- Proper Documentation
Motivation and Context
The existing Inspector is hard to maintain for a number of reasons. We decided to create a more maintainable Inspector from the ground up, that retains the deterministic boundary of operation, but exposes a plugin architecture.
Track Protocol Conformance
- Identify any gaps in V1 inspector
- Track which features are covered
- Implement to parity and fill gaps
- Provide conformance testing in CI
Improve Maintainability
- Clearer separation of concerns
- Smaller, reusable components
Provide Extensibility
- Plugin SDK / API
- Third parties cand develop/maintain their own plugins for features such as
- LLM usage
- Evals
- Multiple server connections
- Apps playground
- Testing of registry configuration
- Interface with registry
Proper Documentation
- One big README won't cut it anymore
- Published documentation in MDX should
- Render in style of spec
- Kept up to date
- Cover all features
- Discuss troubleshooting strategies
- Provide detailed contribution guidance that explains architecture and expected idioms